CAUSAL DISCOVERY FOR IDENTIFYING HIDDEN DEPENDENCIES IN AI-DRIVEN PUBLIC POLICY RECOMMENDATIONS

Artificial Intelligence (AI) has emerged as an important decision-support technology in public policy formulation, enabling governments to analyze complex socioeconomic conditions, predict policy outcomes, and optimize resource allocation. However, conventional machine learning approaches primarily identify statistical associations and may overlook hidden causal dependencies among policy variables, institutional characteristics, and socioeconomic outcomes. Such limitations can produce misleading recommendations, particularly when confounding factors, indirect relationships, and interdependent policy interventions influence decision-making. This study proposes an integrated causal discovery framework for identifying hidden dependencies in AI-driven public policy recommendations by combining causal graphical modelling, constraint-based discovery, score-based learning, and explainable machine learning techniques. The proposed framework is designed to distinguish direct and indirect dependencies, identify potential confounding relationships, and improve the transparency, reliability, and interpretability of policy recommendations. The research adopts a quantitative comparative design using publicly available socioeconomic and governance indicators, with potential applications in healthcare expenditure, educational development, employment, poverty reduction, and public resource allocation. The evaluation framework considers ten numerical comparison tables and three figures covering dataset characteristics, causal dependency identification, predictive accuracy, false-positive rates, structural consistency, policy recommendation reliability, computational efficiency, and decision-support performance. Established approaches, including correlation-based analysis, Random Forest, Bayesian networks, PC-based causal discovery, and NOTEARS, are considered as comparative methods. Performance assessment is structured around standard measures such as precision, recall, F1-score, structural Hamming distance, and computational execution time. The study also incorporates sensitivity analysis to examine the stability of discovered relationships under changes in sample size, variable selection, and confounding conditions. The proposed approach provides a systematic foundation for identifying potentially misleading statistical associations and revealing hidden dependencies that conventional predictive models may fail to capture. By integrating causal discovery with interpretable policy evaluation, the framework aims to strengthen evidence-based governance, improve policy targeting, reduce unintended intervention effects, and support more accountable public decision-making. The study contributes a reproducible evaluation methodology for assessing the potential benefits and limitations of causal AI in complex public policy environments.

Authors

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-11-16
DOI
https://doi.org/10.5281/zenodo.24273760
Primary Topic
Bayesian Modeling and Causal Inference
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

CAUSAL DISCOVERY FOR IDENTIFYING HIDDEN DEPENDENCIES IN AI-DRIVEN PUBLIC POLICY RECOMMENDATIONS

Divya Vishwanath
2 links
Zenodo (CERN European Organization for Nuclear Research)
Bayesian Modeling and Causal Inference
article

CAUSAL DISCOVERY FOR IDENTIFYING HIDDEN DEPENDENCIES IN AI-DRIVEN PUBLIC POLICY RECOMMENDATIONS

Divya Vishwanath
article en

Abstract

Artificial Intelligence (AI) has emerged as an important decision-support technology in public policy formulation, enabling governments to analyze complex socioeconomic conditions, predict policy outcomes, and optimize resource allocation. However, conventional machine learning approaches primarily identify statistical associations and may overlook hidden causal dependencies among policy variables, institutional characteristics, and socioeconomic outcomes. Such limitations can produce misleading recommendations, particularly when confounding factors, indirect relationships, and interdependent policy interventions influence decision-making. This study proposes an integrated causal discovery framework for identifying hidden dependencies in AI-driven public policy recommendations by combining causal graphical modelling, constraint-based discovery, score-based learning, and explainable machine learning techniques. The proposed framework is designed to distinguish direct and indirect dependencies, identify potential confounding relationships, and improve the transparency, reliability, and interpretability of policy recommendations. The research adopts a quantitative comparative design using publicly available socioeconomic and governance indicators, with potential applications in healthcare expenditure, educational development, employment, poverty reduction, and public resource allocation. The evaluation framework considers ten numerical comparison tables and three figures covering dataset characteristics, causal dependency identification, predictive accuracy, false-positive rates, structural consistency, policy recommendation reliability, computational efficiency, and decision-support performance. Established approaches, including correlation-based analysis, Random Forest, Bayesian networks, PC-based causal discovery, and NOTEARS, are considered as comparative methods. Performance assessment is structured around standard measures such as precision, recall, F1-score, structural Hamming distance, and computational execution time. The study also incorporates sensitivity analysis to examine the stability of discovered relationships under changes in sample size, variable selection, and confounding conditions. The proposed approach provides a systematic foundation for identifying potentially misleading statistical associations and revealing hidden dependencies that conventional predictive models may fail to capture. By integrating causal discovery with interpretable policy evaluation, the framework aims to strengthen evidence-based governance, improve policy targeting, reduce unintended intervention effects, and support more accountable public decision-making. The study contributes a reproducible evaluation methodology for assessing the potential benefits and limitations of causal AI in complex public policy environments.

Zenodo (CERN European Organization for Nuclear Research)
Peace, justice, and strong institutions
Bayesian Modeling and Causal Inference
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon